Post Claiming "AI Speeds Up AI Research" Hits 76 Points on Hacker News
A blog post arguing that putting AI to work on research accelerates AI research itself landed on Hacker News, where it drew 76 points and 30 comments.
A post arguing that using AI in research speeds up AI research itself was shared on Hacker News on November 6. It ran on the blog Planned Obsolescence. The post discussed a METR report that examined Anthropic. According to the report, the pace at which Anthropic’s AI capabilities improved from October 2025 to October 2026 was roughly 2x the pace of the preceding 12 months. That pace was measured using AECI, an internal version of the Epoch Capabilities Index.
On Hacker News, the post drew 76 points and 30 comments. The discussion centered on how much research actually speeds up once AI is enlisted to do it. The focus is not AI that helps ordinary users with their work, but the role AI plays in the process of building AI.
To gauge the size of the claim, it needs to be clear which stage of research got faster, and by how much. It also matters what yardstick was used to measure the speedup. Only by weighing the time saved on research alongside the quality of the research output can the claim be judged for what it really amounts to.